Reproduce “FedL2P: Federated Learning to Personalize” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Royson Lee, Minyoung Kim, Da Li, Xinchi Qiu, Timothy Hospedales, Ferenc Huszár, Nicholas D. Lane
FedL2P learns client-specific fine-tuning settings from local data profiles in federated learning. Two meta-networks select batch-normalization statistics and layer-wise learning rates, allowing clients to adapt a shared pretrained model without sharing raw data; clients still fine-tune locally. On CIFAR-10 with 1,000 clients and α=1.0, it achieved 65.76 ± 0.31% test accuracy versus 61.42 ± 0.13% for fixed client statistics, under that protocol.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.